Three title normalization fixes

1. @misc colon subtitle: use only the part after ':' when matching web page titles
2. LaTeX escapes: normalize \$ \% \& \# to their plain characters
3. Test corrections and additions for both cases

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: monperrus

check_hallucinations/main.py

@@ -32,6 +32,10 @@ def _normalize_title(title):
         title.lower()
         .strip()
         .rstrip(".")
+        .replace("\\$", "$")
+        .replace("\\%", "%")
+        .replace("\\&", "&")
+        .replace("\\#", "#")
         .replace(": ", ". ")
         .replace(",", " ")
         .replace("   ", " ")
@@ -93,7 +97,9 @@ def check_misc_url(url, bib_title):
     page_title = _fetch_page_title(url)
     if not page_title:
         raise MiscEntryError(f"No <title> found at {url}")
-    norm_bib = _normalize_title(bib_title)
+    # When the bib title has a colon, the part after it is the real title
+    search_title = bib_title.split(":", 1)[1].strip() if ":" in bib_title else bib_title
+    norm_bib = _normalize_title(search_title)
     norm_page = _normalize_title(page_title)
     if norm_bib not in norm_page:
         raise MiscEntryError(

test_misc.py

@@ -35,3 +35,23 @@ def test_check_misc_url_no_title(mock_fetch):
     mock_fetch.return_value = ""
     with pytest.raises(MiscEntryError, match="No <title>"):
         check_misc_url("https://example.com", "Some Title")
+
+
+@patch("check_hallucinations.main._fetch_page_title")
+def test_check_misc_url_colon_uses_subtitle(mock_fetch):
+    # Page shows only the subtitle; the "Main: " prefix is not in the HTML title
+    mock_fetch.return_value = "The Official Leaderboard | SWE-bench"
+    check_misc_url(
+        "https://www.swebench.com",
+        "SWE-bench: The Official Leaderboard",
+    )
+
+
+@patch("check_hallucinations.main._fetch_page_title")
+def test_check_misc_url_colon_wrong_subtitle(mock_fetch):
+    mock_fetch.return_value = "Some Unrelated Page Title"
+    with pytest.raises(MiscEntryError, match="Title does not match"):
+        check_misc_url(
+            "https://example.com",
+            "My Blog: Adventures in Python",
+        )

test_normalize.py

@@ -18,3 +18,9 @@ def test_diacritics():
 
 def test_curly_braces():
     assert _normalize_title("{Foo} Bar") == _normalize_title("Foo Bar")
+
+
+def test_latex_dollar_escape():
+    bib = r"Fixing 55 out of 105 bugs for \$8 each"
+    ss  = "Fixing 55 out of 105 bugs for $8 each"
+    assert _normalize_title(bib) == _normalize_title(ss)